Monday, January 10, 2011

4 systems, built same way, patch outcome different

Built a clustered system, 4 nodes, with Solaris 10 months ago. Finally hit enough problems that it's decided that finally we get our patch upgrade. zfs root rocks again, lucreate, lumount, pca, luactivate, init 6 and we're on the patched OS.

1) system 1 - /usr/java still pointing at jdk/jdk_1.6.0_14, appearently patch install failed. logs disagree... Reinstall patches and fix /usr/java *again*.

2) system 2 - /usr/java now pointing at jdk/jdk_1.5.0_27, huh? 1.5 now? Fix link.

3) system 3 - /usr/java now pointing at jdk/jdk_1.6.0_23, expected

4) system 4 - /usr/java now pointing at jdk/jdk_1.5.0_27. sheesh

You'd expect that 4 systems, built exactly the same way, patched exactly the same way, would have the same outcome from *deterministic* code from the patch install. The law on unintended consequences strikes again.

No comments:

Post a Comment